Unshakable Pacifism?

A psychometric approach to Japan’s pacifism and antimilitarism

Jozef Rivest
jrivest@umich.edu

Department of Political Science, University of Michigan

The Debate and Its Stakes

  • Security environment has deteriorated; Trump and the threats of removing the bases has eroded confidence in the alliance; Abe’s reforms (2012–2020) enacted significant policy changes — Berger (1993)’s two conditions, and more, for antimilitarist culture erosion are now met

From Norms to Attitudes

  • Previous research: pacifism/antimilitarism as macro-level norms — collective, national, homogenizing

  • This paper: individual-level attitudes — unobservable latent variables, continuously distributed, shaped by personal experience

  • This reframing reveals heterogeneity that norm-based approaches obscure

Three Untested Assumptions

  1. Unidimensionality — all defense/security items measure a single construct; any item is interchangeable

  2. Equal item contribution — all items contribute equally; any single item is a sufficient proxy

  3. Measurement equivalence — items measure the same construct across demographic groups, enabling valid cross-group comparisons

If these fail, existing findings about pacifist change may rest on invalid measurement

Data and Methods

Data and Methods

  • Data: 2022 UTokyo-Asahi Survey (UTAS), n=1,014 (listwise deletion)
    • 9 items on defense, security, and legal constraints
    • Excludes items asking preference between two policy options
  • Three-stage psychometric strategy:
    1. EFA — determine dimensional structure
    2. IRT (GRM) — examine item contribution and information
    3. Measurement invariance + DIF — test equivalence across gender and age cohorts

Questions

Label Wording
Article 9 The Liberal Democratic Party is proposing a constitutional amendment bill that would leave Paragraphs 1 and 2 of Article 9 unchanged, but would newly stipulate the existence of the Self-Defense Forces. Do you agree with these amendments to Article 9? Are you against it?
Change constitution Do you think it is necessary to change the current constitution? Do you think there is no need to change?
Preemptive capabilities Japan's Self-Defense Forces do not have the ability to attack enemy missile bases. Because attacks on enemy bases depend on the American military. Do you support Japan having the ability to attack enemy missile bases? Are you against it?
Alliance USA Are you in favor of continuing to maintain the Japan-U.S. Security Treaty? Are you against it?
Futenma Are you in favor of relocating the U.S. military's Futenma Air Base in Okinawa Prefecture to Henoko, Nago City, Okinawa Prefecture? Are you against it?
Collective SDF If a country with close ties to Japan is attacked by another country, even if Japan is not attacked, the Self-Defense Forces will be able to join forces only if the government determines that there is a risk that Japan's existence is threatened. You can fight. This is called the limited exercise of the right of collective self-defense. In the unlikely event that a war breaks out between the United States and a country surrounding Japan, what do you think should be done regarding the exercise of this right of collective self-defense?
Defense Power Japan's defense power should be further strengthened
Three Non Nuclear Japan has three non-nuclear principles: ``Do not possess,'' ``Do not create,'' and ``Do not allow nuclear weapons to be brought into Japan.'' What do you think should be done about these three non-nuclear principles?
Yasukuni I want the Prime Minister to visit Yasukuni Shrine

Analyses

Factor Structure: Two Dimensions

Relationship Between Items and Latent Factors

Item Contribution to Measurement

Contribution of the test as a whole

Measurement Invariance

Across age cohorts:

  • Full strong invariance holds — loadings and intercepts equivalent across generations
  • Valid basis for cross-generational comparisons

Across gender:

  • Configural and weak invariance hold
  • Strong invariance marginally fails (p = 0.090) — factor loadings approach non-equivalence
  • Item intercepts remain fully equivalent
  • Cross-gender comparisons should be made with caution

Differential Item Functioning

Gender (PA1 — domestic):

  • Change Constitution and Yasukuni flagged — women discriminate more sharply on Change Constitution (non-uniform + uniform DIF)
  • Effect sizes negligible by Zumbo-Thomas / Jodoin-Gierl criteria

Gender (PA2 — international):

  • Article 9 and Futenma flagged — women more likely to oppose Futenma relocation at the same level of θ₂

Age cohorts (PA1):

  • Preemptive Capabilities and Change Constitution — War generation opposes preemptive strike even at moderate θ₁ (lived memory vs. strategic lens); Postwar cohort requires stronger pacifist orientation
  • No DIF on PA2 — alliance and base politics function equivalently across generations

Discussion

Testing the Three Assumptions

  1. Unidimensionality — Rejected
    • Two dimensions: domestic capabilities (nuclear taboo, preemptive strike, constitutional revision, defense spending) vs. international/alliance (U.S. alliance, base politics)
  2. Equal item contribution — Rejected
    • Items vary substantially in discrimination; Three Non-Nuclear discriminates strongly on PA1; Yasukuni, Article 9, and Collective SDF the weakest
  3. Measurement equivalence — Partial
    • Full strong invariance across age cohorts; borderline across gender (p = 0.090)
    • DIF: Change Constitution and Yasukuni gender-biased on PA1; Futenma and Article 9 on PA2; Preemptive Capabilities generationally differentiated on PA1

Implications

  • The two-dimensional structure explains contradictory findings in the literature: different groups may be changing on different dimensions

    • Robustness tests in the supplementary materials

  • Japanese respondents can simultaneously support domestic defense within constitutional bounds while relying on the U.S. alliance — a nuance lost under unidimensionality

  • Single-item inferences are doubly unreliable: they capture only one dimension and may be biased for specific groups

    • They also have unequal capacity to capture and measure the latent variables of interest

Conclusion and Limitations

What Do We Do Now?

  • A validated multi-item scale — accounting for dimensionality and unequal item contribution — offers a more robust basis for inference

  • Be more careful about inferences made from individual items

  • Future: Work on a battery of items that would provide a reliable and consistent way to measure these attitudes

    • Open-ended survey questions + LLMs to examine how people define these concepts → develop better items

Limitations

  • Cross-sectional data (2022) — a snapshot, not temporal dynamics

  • Does not provide a definitive answer to whether pacifism has changed

    • Aldrich-McKelvey scale to estimate variation over time and correct for DIF (Hare et al. 2015)

  • Dimensional structure over time remains unknown

  • At most: a methodological benchmark for future longitudinal research

Bibliography

Berger, Thomas U. 1993. “From Sword to Chrysanthemum: Japan’s Culture of Anti-Militarism.” International Security 17 (4): 119–50. https://doi.org/10.2307/2539024.
Gustafsson, Karl. 2020. “Temporal Othering, de-Securitisation and Apologies: Understanding Japanese Security Policy Change.” Journal of International Relations and Development 23 (3): 511–34. https://doi.org/10.1057/s41268-018-00168-y.
Hare, Christopher, David A. Armstrong, Ryan Bakker, Royce Carroll, and Keith T. Poole. 2015. “Using Bayesian Aldrich-McKelvey Scaling to Study Citizens’ Ideological Preferences and Perceptions.” American Journal of Political Science 59 (3): 759–74. https://doi.org/10.1111/ajps.12151.
Hughes, Christopher W. 2022. “Remilitarization in Japan.” In The Oxford Handbook of Japanese Politics, edited by Robert J. Pekkanen and Saadia M. Pekkanen, 680–700. Oxford University Press. https://doi.org/10.1093/oxfordhb/9780190050993.013.36.
Katzenstein, Peter J., and Nobuo Okawara. 1993. Japan’s National Security. Structures, Norms and Policy Responses in a Changing World. Ithaca, New York: Cornell East Asia Series.
Kim, Ji Young. 2015. “Dismantling the Final Barrier: Transforming Japan into a ‘Normal Country’ in the Post-Cold-War Era.” Pacific Focus 30 (2): 223–48. https://doi.org/10.1111/pafo.12049.
Le, Tom Phuong. 2021. Japan’s Aging Peace. Pacificism and Militarism in the Twenty-First Century. New York: Columbia University Press.
Liff, Adam P., and Ko Maeda. 2019. “Electoral Incentives, Policy Compromise, and Coalition Durability: Japan’s LDPKomeito Government in a Mixed Electoral System.” Japanese Journal of Political Science 20 (1): 53–73. https://doi.org/10.1017/S1468109918000415.
Lind, Jennifer. 2004. “Pacifism or Passing the Buck? Testing Theories of Japanese Security Policy.” International Security 29 (1): 92–121. https://doi.org/10.1162/0162288041762968.
Lind, Jennifer, and Chikako Kawakatsu Ueki. 2021. “IS JAPAN BACK? MEASURING NATIONALISM AND MILITARY ASSERTIVENESS IN ASIA’S OTHER GREAT POWER.” Journal of East Asian Studies 21 (3): 367–401. https://doi.org/10.1017/jea.2021.20.
Smith, Sheila. 2019. Japan Rearmed. Cambridge, Massachusetts: Harvard University Press.
Takenaka, Harutaka. 2021. “Evolution of Japanese Security Policy and the House of Councilors.” Japanese Journal of Political Science 22: 96–115. https://doi.org/doi:10.1017/S1468109921000086.

Appendix

Scree plot

Factor Analysis

Items PA1 PA2 Communality Uniqueness Complexity
Article 9 0.392 0.33 0.408 0.592 1.94
Three Non Nuclear 1.022 -0.263 0.81 0.19 1.13
Preemptive capabilities 0.842 -0.102 0.622 0.378 1.03
Alliance USA -0.352 1.025 0.766 0.234 1.23
Futenma -0.161 -0.511 0.38 0.62 1.2
Collective SDF 0.524 0.195 0.428 0.572 1.27
Change Constitution 0.723 -0.083 0.461 0.539 1.03
Defense Power 0.745 0.07 0.619 0.381 1.02
Yasukuni 0.3 0.212 0.207 0.793 1.8
Information
SS loadings 3.3 1.4
Proportion Var 0.37 0.16
Proportion Explained 0.7 0.3
Correlation between factors
PA1 1 0.57
PA2 0.57 1
GOF Metrics
Correlation of (regression) scores with factors 1 1
Multiple R square of scores with factors 1 1
Minimum correlation of possible factor scores 1 1

IRT: Discrimination

Items Discrimination 1 Discrimination 2
Article 9 0.9 0.734
Three Non Nuclear 3.436 -0.661
Preemptive capabilities 2.698 -0.517
Alliance USA -1.055 2.99
Futenma -0.333 -1.095
Collective SDF 1.235 0.473
Change Constitution 1.634 -0.261
Defense Power 2.302 0.148
Yasukuni 0.626 0.44

IRT: Difficulty

Items Discrimination Difficulty 1 Difficulty 2 Difficulty 3 Difficulty 4
Article 9 1.47 0.411 - - -
Three Non Nuclear 3.08 -0.829 - - -
Preemptive capabilities 2.42 0.052 - - -
Alliance USA 2.49 1.395 - - -
Futenma 1.32 -0.097 - - -
Collective SDF 1.57 -0.934 0.58 2.1 -
Change Constitution 1.49 0.366 - - -
Defense Power 2.39 -0.441 0.5 1.5 2
Yasukuni 0.96 -1.413 -0.51 1.7 2.3

DIF: Gender and Base Politics

DIF: Generational Memory and Preemptive Strike